Grok 4.5 is xAI's current flagship model, available through the Grok app, X, and the API under the model name grok-4.5. It launched in July 2026 with a 500,000-token context window and API pricing of $2 per million input tokens and $6 per million output tokens for prompts under 200K tokens.
Short answer: Grok 4.5 is xAI's flagship model as of July 2026, with a 500K-token context window. API pricing is $2/M input and $6/M output tokens under 200K tokens (doubling above that threshold), and it scores 54 on the Artificial Analysis Intelligence Index, ranking #13 of 190 models tested. It's accessible through the Grok app, X Premium, or the xAI API.
In my testing, I pulled the numbers below straight from xAI’s own docs and cross-checked them against Artificial Analysis’s independent benchmark run rather than repeating whatever a pricing tracker claimed, because the model-name churn in this space (4.3, 4.20, 4.5, Heavy) makes it easy to quote the wrong tier by accident. Here's what the model actually costs, how it stacks up against the other current flagships, and the exact steps to start using it, whether you're chatting in the app or calling the API.
What you'll need
For the consumer app, you just need a Grok or X account — Grok has a free tier with basic chat, though it's staged for model access, so a free account isn't guaranteed to show it in the picker the day it ships. I couldn't get a clean read of xAI's own pricing page while researching this (it returned a bot-check page every time I tried), so if you want the current SuperGrok or SuperGrok Heavy subscription price straight from the source, check x.ai/pricing yourself — I'm not going to publish a subscription number I couldn't verify firsthand. For the API path, you need an account at console.x.ai, a payment method on file, and a budget in mind: at $2/M input tokens, a 10,000-word document (roughly 13,000 tokens) costs about 3 cents to send once.
Step-by-step: using Grok 4.5
1. Decide between the app and the API
If you just want to chat, use the Grok app or X — no code required, and the model picker shows which version you're talking to. If you're building something, or you want to feed it a long document, the API gives you the full 500K-token context window and predictable per-token billing.
2. Check the model picker before you trust the answer
Grok's rollout to 4.5 has been staged across tiers rather than instant for every account. Open the model picker in the app and confirm it says "4.5" — if it still shows 4.3 or 4.20, your account hasn't been switched over yet, and any benchmark claims you've read about 4.5 won't apply to what you're actually running.
3. Get an API key if you're building anything
Sign in at console.x.ai, generate a key, and call the model as grok-4.5 in your request. xAI's docs list a February 1, 2026 knowledge cutoff for the model, so anything more recent needs a search or browsing tool attached to the request — the base model won't know it.
4. Watch the 200K-token pricing cliff
It bills at $2/M input and $6/M output under 200K prompt tokens, but the whole request jumps to $4/M input and $12/M output once you cross that line — not just the tokens over the threshold. If you're sending long documents, check your token count before you submit, not after the bill arrives.
5. Turn on caching for repeated context
Cached input tokens run $0.30/M under the 200K threshold, a 6x discount versus fresh input. If you're running the same system prompt or reference document across many calls, caching is where the real savings are, not shrinking your prompts.
Example prompts you can copy
These lean on what the model is actually built for — a large context window and current-events grounding when search is enabled — rather than generic chat requests:
- "Here's a 40-page contract [paste text]. Flag every clause that references a deadline or penalty, and list them in order." (Tests the 500K context window on a real document instead of a toy example.)
- "Search for the latest coverage of [topic] and summarize what changed in the last 48 hours, citing your sources." (Uses search grounding to work around the February 2026 training cutoff.)
- "Write a Python function to [task], then write three unit tests that would catch a subtle off-by-one bug in it." (Coding plus self-critique in one pass.)
- "Compare these two API responses for factual accuracy: [paste both]." (Useful once you're running it against another model's output side by side.)
- "Summarize this thread in three sentences, then tell me what's missing from the argument." (Good for testing reasoning depth, not just summarization.)
Common mistakes to avoid
The first mistake I'd flag: assuming your SuperGrok or Grok app subscription already has 4.5 the day it's announced — the rollout is staged, and the model picker is the only reliable way to check, not the release-date headline. Second, quoting the $2/$6 per-million rate for every request; it only holds under 200K prompt tokens, and xAI doubles the rate above that line rather than pro-rating it. Third, expecting it to know about anything after its February 1, 2026 training cutoff without enabling a search or browsing tool — the base model will confidently guess instead of telling you it doesn't know. Fourth, comparing the Intelligence Index score in isolation; Artificial Analysis measured its output speed at 56.4 tokens per second, which is below average for a model in its class, so a strong reasoning score doesn't mean it's the fastest option for latency-sensitive work. Fifth, mixing up grok-4.5 with grok-4.3 or grok-4.20 in API calls — they're priced differently and one has a much smaller context window, so a typo in the model string changes your bill.
Grok 4.5 vs. the other current flagships
| Model | Input $/M tokens | Output $/M tokens | Context window | Notable |
|---|---|---|---|---|
| Grok 4.5 | $2.00 (under 200K) | $6.00 (under 200K) | 500K tokens | Rank #13 of 190 on Artificial Analysis Intelligence Index (score 54) |
| Claude Sonnet 5 | $2.00 (intro, through Aug 31, 2026) | $10.00 (intro) | 1M tokens | Rises to $3/$15 per M on Sept 1, 2026 |
| GPT-5.6 Sol | $5.00 | $30.00 | ~1.05M tokens | OpenAI's flagship for complex reasoning and coding |
Prices verified against each vendor's own pricing docs on 2026-07-27; its rates double for prompts at or above 200K tokens, and the other two models don't currently have that cliff.
Tools that make this easier
If you're deciding whether it's worth adding alongside the model you already use, my how to use ChatGPT, how to use Claude AI, and how to use Gemini guides cover the setup quirks specific to each of those, so you know what you're actually comparing it against. If coding is your main use case, best AI tool for code tests Grok's coding performance against Claude Code and Cursor head-to-head, and ChatGPT alternatives for coding covers the wider field if none of the big three fit. My starter kit for AI is the place to start if you haven't set up any of these accounts yet. And for how I verify numbers like the ones in this piece before publishing them, see how we test AI tools.
Where this leaves you
Grok 4.5 is a genuinely capable model on paper — a half-million-token context window and a top-15 Intelligence Index score are nothing to dismiss — but it isn't the fastest or the cheapest option in its tier, and the per-model pricing cliff at 200K tokens is easy to miss if you're not watching your token counts. If you already pay for X Premium or SuperGrok, checking whether your model picker shows 4.5 costs nothing. If you're building on the API, the math is straightforward: cheaper per token than GPT-5.6 Sol, more expensive than Claude Sonnet 5's introductory rate, with the largest context window of the three only if you stay under the 200K threshold.
Last updated: July 27, 2026 · By Vishal Swami, Founder & Lead AI Reviewer, AISagely
Frequently Asked Questions
Grok 4.5: is it free?
The Grok app has a free tier, but full access to the 4.5 model is gated behind paid tiers as the rollout stages out. The API isn't free — it's pay-per-token at $2/M input and $6/M output under 200K tokens, billed to a card on file at console.x.ai.
How long does it take to start using Grok 4.5?
Through the app, minutes — sign in and check the model picker. Through the API, expect 10-15 minutes to create an account at console.x.ai, generate a key, and send a first test call, plus a few minutes more to confirm your token counts before running anything at scale.
What is the easiest way to try Grok 4.5?
Open the Grok app or X, check whether the model picker already shows 4.5, and run a couple of real prompts — like summarizing a long document — before deciding whether to add the API for anything more serious.
Does Grok 4.5 know about current events?
Not by default. Its training data runs through February 1, 2026, per xAI's own model docs, so anything after that needs a search or browsing tool enabled in the request or app setting.
How does Grok 4.5 compare to Claude Sonnet 5 on price?
They tie on input at $2/M tokens today, but it's cheaper on output at $6/M versus Claude Sonnet 5's introductory $10/M. Claude Sonnet 5's rate rises to $3/$15 per M on September 1, 2026, which makes it the cheaper model on both ends from that date on, as long as you stay under its 200K-token pricing threshold.